In my logistic regression model, I want to remove all the significant variables and keep only the non-significant ones. Because I want to have only the variables that are not distinguished between « presence » and « absence » observations. I have to automatize this procedure in many cases. So I’m thinking to use the opposite of the results given by the Stepwise Algorithms and I want to know if it’s a good way to do this.
Yes, generally using the opposite of the results given by the stepwise algorithms may be a path.
However, by 'opposite' I mean excluding just the first selected variable. Then, exclude this variable from the database.
Repeat the process until there are no longer any largely significant variables as indicated by the low R-Square value.
Still a bit iterative, and not precisely mechanical.